A Watershed Segmentation Method with Shape-Based Merging Criterion Based on Multispectral Remote Sensing Imagery

作者：Yue Anzhi; Yang Jianyu; Zhang Chao^*; Su Wei; Yun Wenju; Zhang Yingna

来源：Sensor Letters, 2011, 9(3): 1006-1013.

DOI：10.1166/sl.2011.1389

• Multispectral Watershed Transformation
• High Resolution Imagery
• Shape Merging Criterion
• Image Segmentation
• Shape Index

Image segmentation is an important preprocessing procedure for image process and application. The result of image segmentation significantly affects the accuracy of subsequent image analysis. Watershed transformation is a powerful morphological tool for image segmentation, which was originally defined for gradient magnitude of gray images. An extension of the watershed segmentation method for multispectral remote sensing imagery was presented in this paper. Firstly, the watershed transformation was applied to each channel respectively. Then, an overlay function was proposed for mixing all initial segments derived from each channel to generate the final mixture segment, which was then input into post segmentation procedure for region merging. Additionally, a new region merging criterion combining region spectral homogeneity and spatial heterogeneity was proposed, in order to reduce the deviation from compactness of region. Finally, the segmentation results were evaluated and compared with other image segmentation, in terms of visual inspection, spectral heterogeneity and shape factor, such as region border index and shape index. The experimental results indicated that the proposed method can produce more visually appropriate segmentation results and higher accurate region boundary.

• 出版日期2011-6
• 单位中国农业大学